xref: /linux/kernel/reboot.c (revision 24bce201d79807b668bf9d9e0aca801c5c0d5f78)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/reboot.c
4  *
5  *  Copyright (C) 2013  Linus Torvalds
6  */
7 
8 #define pr_fmt(fmt)	"reboot: " fmt
9 
10 #include <linux/atomic.h>
11 #include <linux/ctype.h>
12 #include <linux/export.h>
13 #include <linux/kexec.h>
14 #include <linux/kmod.h>
15 #include <linux/kmsg_dump.h>
16 #include <linux/reboot.h>
17 #include <linux/suspend.h>
18 #include <linux/syscalls.h>
19 #include <linux/syscore_ops.h>
20 #include <linux/uaccess.h>
21 
22 /*
23  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
24  */
25 
26 static int C_A_D = 1;
27 struct pid *cad_pid;
28 EXPORT_SYMBOL(cad_pid);
29 
30 #if defined(CONFIG_ARM)
31 #define DEFAULT_REBOOT_MODE		= REBOOT_HARD
32 #else
33 #define DEFAULT_REBOOT_MODE
34 #endif
35 enum reboot_mode reboot_mode DEFAULT_REBOOT_MODE;
36 EXPORT_SYMBOL_GPL(reboot_mode);
37 enum reboot_mode panic_reboot_mode = REBOOT_UNDEFINED;
38 
39 /*
40  * This variable is used privately to keep track of whether or not
41  * reboot_type is still set to its default value (i.e., reboot= hasn't
42  * been set on the command line).  This is needed so that we can
43  * suppress DMI scanning for reboot quirks.  Without it, it's
44  * impossible to override a faulty reboot quirk without recompiling.
45  */
46 int reboot_default = 1;
47 int reboot_cpu;
48 enum reboot_type reboot_type = BOOT_ACPI;
49 int reboot_force;
50 
51 struct sys_off_handler {
52 	struct notifier_block nb;
53 	int (*sys_off_cb)(struct sys_off_data *data);
54 	void *cb_data;
55 	enum sys_off_mode mode;
56 	bool blocking;
57 	void *list;
58 };
59 
60 /*
61  * Temporary stub that prevents linkage failure while we're in process
62  * of removing all uses of legacy pm_power_off() around the kernel.
63  */
64 void __weak (*pm_power_off)(void);
65 
66 /**
67  *	emergency_restart - reboot the system
68  *
69  *	Without shutting down any hardware or taking any locks
70  *	reboot the system.  This is called when we know we are in
71  *	trouble so this is our best effort to reboot.  This is
72  *	safe to call in interrupt context.
73  */
74 void emergency_restart(void)
75 {
76 	kmsg_dump(KMSG_DUMP_EMERG);
77 	machine_emergency_restart();
78 }
79 EXPORT_SYMBOL_GPL(emergency_restart);
80 
81 void kernel_restart_prepare(char *cmd)
82 {
83 	blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
84 	system_state = SYSTEM_RESTART;
85 	usermodehelper_disable();
86 	device_shutdown();
87 }
88 
89 /**
90  *	register_reboot_notifier - Register function to be called at reboot time
91  *	@nb: Info about notifier function to be called
92  *
93  *	Registers a function with the list of functions
94  *	to be called at reboot time.
95  *
96  *	Currently always returns zero, as blocking_notifier_chain_register()
97  *	always returns zero.
98  */
99 int register_reboot_notifier(struct notifier_block *nb)
100 {
101 	return blocking_notifier_chain_register(&reboot_notifier_list, nb);
102 }
103 EXPORT_SYMBOL(register_reboot_notifier);
104 
105 /**
106  *	unregister_reboot_notifier - Unregister previously registered reboot notifier
107  *	@nb: Hook to be unregistered
108  *
109  *	Unregisters a previously registered reboot
110  *	notifier function.
111  *
112  *	Returns zero on success, or %-ENOENT on failure.
113  */
114 int unregister_reboot_notifier(struct notifier_block *nb)
115 {
116 	return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
117 }
118 EXPORT_SYMBOL(unregister_reboot_notifier);
119 
120 static void devm_unregister_reboot_notifier(struct device *dev, void *res)
121 {
122 	WARN_ON(unregister_reboot_notifier(*(struct notifier_block **)res));
123 }
124 
125 int devm_register_reboot_notifier(struct device *dev, struct notifier_block *nb)
126 {
127 	struct notifier_block **rcnb;
128 	int ret;
129 
130 	rcnb = devres_alloc(devm_unregister_reboot_notifier,
131 			    sizeof(*rcnb), GFP_KERNEL);
132 	if (!rcnb)
133 		return -ENOMEM;
134 
135 	ret = register_reboot_notifier(nb);
136 	if (!ret) {
137 		*rcnb = nb;
138 		devres_add(dev, rcnb);
139 	} else {
140 		devres_free(rcnb);
141 	}
142 
143 	return ret;
144 }
145 EXPORT_SYMBOL(devm_register_reboot_notifier);
146 
147 /*
148  *	Notifier list for kernel code which wants to be called
149  *	to restart the system.
150  */
151 static ATOMIC_NOTIFIER_HEAD(restart_handler_list);
152 
153 /**
154  *	register_restart_handler - Register function to be called to reset
155  *				   the system
156  *	@nb: Info about handler function to be called
157  *	@nb->priority:	Handler priority. Handlers should follow the
158  *			following guidelines for setting priorities.
159  *			0:	Restart handler of last resort,
160  *				with limited restart capabilities
161  *			128:	Default restart handler; use if no other
162  *				restart handler is expected to be available,
163  *				and/or if restart functionality is
164  *				sufficient to restart the entire system
165  *			255:	Highest priority restart handler, will
166  *				preempt all other restart handlers
167  *
168  *	Registers a function with code to be called to restart the
169  *	system.
170  *
171  *	Registered functions will be called from machine_restart as last
172  *	step of the restart sequence (if the architecture specific
173  *	machine_restart function calls do_kernel_restart - see below
174  *	for details).
175  *	Registered functions are expected to restart the system immediately.
176  *	If more than one function is registered, the restart handler priority
177  *	selects which function will be called first.
178  *
179  *	Restart handlers are expected to be registered from non-architecture
180  *	code, typically from drivers. A typical use case would be a system
181  *	where restart functionality is provided through a watchdog. Multiple
182  *	restart handlers may exist; for example, one restart handler might
183  *	restart the entire system, while another only restarts the CPU.
184  *	In such cases, the restart handler which only restarts part of the
185  *	hardware is expected to register with low priority to ensure that
186  *	it only runs if no other means to restart the system is available.
187  *
188  *	Currently always returns zero, as atomic_notifier_chain_register()
189  *	always returns zero.
190  */
191 int register_restart_handler(struct notifier_block *nb)
192 {
193 	return atomic_notifier_chain_register(&restart_handler_list, nb);
194 }
195 EXPORT_SYMBOL(register_restart_handler);
196 
197 /**
198  *	unregister_restart_handler - Unregister previously registered
199  *				     restart handler
200  *	@nb: Hook to be unregistered
201  *
202  *	Unregisters a previously registered restart handler function.
203  *
204  *	Returns zero on success, or %-ENOENT on failure.
205  */
206 int unregister_restart_handler(struct notifier_block *nb)
207 {
208 	return atomic_notifier_chain_unregister(&restart_handler_list, nb);
209 }
210 EXPORT_SYMBOL(unregister_restart_handler);
211 
212 /**
213  *	do_kernel_restart - Execute kernel restart handler call chain
214  *
215  *	Calls functions registered with register_restart_handler.
216  *
217  *	Expected to be called from machine_restart as last step of the restart
218  *	sequence.
219  *
220  *	Restarts the system immediately if a restart handler function has been
221  *	registered. Otherwise does nothing.
222  */
223 void do_kernel_restart(char *cmd)
224 {
225 	atomic_notifier_call_chain(&restart_handler_list, reboot_mode, cmd);
226 }
227 
228 void migrate_to_reboot_cpu(void)
229 {
230 	/* The boot cpu is always logical cpu 0 */
231 	int cpu = reboot_cpu;
232 
233 	cpu_hotplug_disable();
234 
235 	/* Make certain the cpu I'm about to reboot on is online */
236 	if (!cpu_online(cpu))
237 		cpu = cpumask_first(cpu_online_mask);
238 
239 	/* Prevent races with other tasks migrating this task */
240 	current->flags |= PF_NO_SETAFFINITY;
241 
242 	/* Make certain I only run on the appropriate processor */
243 	set_cpus_allowed_ptr(current, cpumask_of(cpu));
244 }
245 
246 /**
247  *	kernel_restart - reboot the system
248  *	@cmd: pointer to buffer containing command to execute for restart
249  *		or %NULL
250  *
251  *	Shutdown everything and perform a clean reboot.
252  *	This is not safe to call in interrupt context.
253  */
254 void kernel_restart(char *cmd)
255 {
256 	kernel_restart_prepare(cmd);
257 	migrate_to_reboot_cpu();
258 	syscore_shutdown();
259 	if (!cmd)
260 		pr_emerg("Restarting system\n");
261 	else
262 		pr_emerg("Restarting system with command '%s'\n", cmd);
263 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
264 	machine_restart(cmd);
265 }
266 EXPORT_SYMBOL_GPL(kernel_restart);
267 
268 static void kernel_shutdown_prepare(enum system_states state)
269 {
270 	blocking_notifier_call_chain(&reboot_notifier_list,
271 		(state == SYSTEM_HALT) ? SYS_HALT : SYS_POWER_OFF, NULL);
272 	system_state = state;
273 	usermodehelper_disable();
274 	device_shutdown();
275 }
276 /**
277  *	kernel_halt - halt the system
278  *
279  *	Shutdown everything and perform a clean system halt.
280  */
281 void kernel_halt(void)
282 {
283 	kernel_shutdown_prepare(SYSTEM_HALT);
284 	migrate_to_reboot_cpu();
285 	syscore_shutdown();
286 	pr_emerg("System halted\n");
287 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
288 	machine_halt();
289 }
290 EXPORT_SYMBOL_GPL(kernel_halt);
291 
292 /*
293  *	Notifier list for kernel code which wants to be called
294  *	to prepare system for power off.
295  */
296 static BLOCKING_NOTIFIER_HEAD(power_off_prep_handler_list);
297 
298 /*
299  *	Notifier list for kernel code which wants to be called
300  *	to power off system.
301  */
302 static ATOMIC_NOTIFIER_HEAD(power_off_handler_list);
303 
304 static int sys_off_notify(struct notifier_block *nb,
305 			  unsigned long mode, void *cmd)
306 {
307 	struct sys_off_handler *handler;
308 	struct sys_off_data data = {};
309 
310 	handler = container_of(nb, struct sys_off_handler, nb);
311 	data.cb_data = handler->cb_data;
312 	data.mode = mode;
313 	data.cmd = cmd;
314 
315 	return handler->sys_off_cb(&data);
316 }
317 
318 static struct sys_off_handler platform_sys_off_handler;
319 
320 static struct sys_off_handler *alloc_sys_off_handler(int priority)
321 {
322 	struct sys_off_handler *handler;
323 	gfp_t flags;
324 
325 	/*
326 	 * Platforms like m68k can't allocate sys_off handler dynamically
327 	 * at the early boot time because memory allocator isn't available yet.
328 	 */
329 	if (priority == SYS_OFF_PRIO_PLATFORM) {
330 		handler = &platform_sys_off_handler;
331 		if (handler->cb_data)
332 			return ERR_PTR(-EBUSY);
333 	} else {
334 		if (system_state > SYSTEM_RUNNING)
335 			flags = GFP_ATOMIC;
336 		else
337 			flags = GFP_KERNEL;
338 
339 		handler = kzalloc(sizeof(*handler), flags);
340 		if (!handler)
341 			return ERR_PTR(-ENOMEM);
342 	}
343 
344 	return handler;
345 }
346 
347 static void free_sys_off_handler(struct sys_off_handler *handler)
348 {
349 	if (handler == &platform_sys_off_handler)
350 		memset(handler, 0, sizeof(*handler));
351 	else
352 		kfree(handler);
353 }
354 
355 /**
356  *	register_sys_off_handler - Register sys-off handler
357  *	@mode: Sys-off mode
358  *	@priority: Handler priority
359  *	@callback: Callback function
360  *	@cb_data: Callback argument
361  *
362  *	Registers system power-off or restart handler that will be invoked
363  *	at the step corresponding to the given sys-off mode. Handler's callback
364  *	should return NOTIFY_DONE to permit execution of the next handler in
365  *	the call chain or NOTIFY_STOP to break the chain (in error case for
366  *	example).
367  *
368  *	Multiple handlers can be registered at the default priority level.
369  *
370  *	Only one handler can be registered at the non-default priority level,
371  *	otherwise ERR_PTR(-EBUSY) is returned.
372  *
373  *	Returns a new instance of struct sys_off_handler on success, or
374  *	an ERR_PTR()-encoded error code otherwise.
375  */
376 struct sys_off_handler *
377 register_sys_off_handler(enum sys_off_mode mode,
378 			 int priority,
379 			 int (*callback)(struct sys_off_data *data),
380 			 void *cb_data)
381 {
382 	struct sys_off_handler *handler;
383 	int err;
384 
385 	handler = alloc_sys_off_handler(priority);
386 	if (IS_ERR(handler))
387 		return handler;
388 
389 	switch (mode) {
390 	case SYS_OFF_MODE_POWER_OFF_PREPARE:
391 		handler->list = &power_off_prep_handler_list;
392 		handler->blocking = true;
393 		break;
394 
395 	case SYS_OFF_MODE_POWER_OFF:
396 		handler->list = &power_off_handler_list;
397 		break;
398 
399 	case SYS_OFF_MODE_RESTART:
400 		handler->list = &restart_handler_list;
401 		break;
402 
403 	default:
404 		free_sys_off_handler(handler);
405 		return ERR_PTR(-EINVAL);
406 	}
407 
408 	handler->nb.notifier_call = sys_off_notify;
409 	handler->nb.priority = priority;
410 	handler->sys_off_cb = callback;
411 	handler->cb_data = cb_data;
412 	handler->mode = mode;
413 
414 	if (handler->blocking) {
415 		if (priority == SYS_OFF_PRIO_DEFAULT)
416 			err = blocking_notifier_chain_register(handler->list,
417 							       &handler->nb);
418 		else
419 			err = blocking_notifier_chain_register_unique_prio(handler->list,
420 									   &handler->nb);
421 	} else {
422 		if (priority == SYS_OFF_PRIO_DEFAULT)
423 			err = atomic_notifier_chain_register(handler->list,
424 							     &handler->nb);
425 		else
426 			err = atomic_notifier_chain_register_unique_prio(handler->list,
427 									 &handler->nb);
428 	}
429 
430 	if (err) {
431 		free_sys_off_handler(handler);
432 		return ERR_PTR(err);
433 	}
434 
435 	return handler;
436 }
437 EXPORT_SYMBOL_GPL(register_sys_off_handler);
438 
439 /**
440  *	unregister_sys_off_handler - Unregister sys-off handler
441  *	@handler: Sys-off handler
442  *
443  *	Unregisters given sys-off handler.
444  */
445 void unregister_sys_off_handler(struct sys_off_handler *handler)
446 {
447 	int err;
448 
449 	if (IS_ERR_OR_NULL(handler))
450 		return;
451 
452 	if (handler->blocking)
453 		err = blocking_notifier_chain_unregister(handler->list,
454 							 &handler->nb);
455 	else
456 		err = atomic_notifier_chain_unregister(handler->list,
457 						       &handler->nb);
458 
459 	/* sanity check, shall never happen */
460 	WARN_ON(err);
461 
462 	free_sys_off_handler(handler);
463 }
464 EXPORT_SYMBOL_GPL(unregister_sys_off_handler);
465 
466 static void devm_unregister_sys_off_handler(void *data)
467 {
468 	struct sys_off_handler *handler = data;
469 
470 	unregister_sys_off_handler(handler);
471 }
472 
473 /**
474  *	devm_register_sys_off_handler - Register sys-off handler
475  *	@dev: Device that registers handler
476  *	@mode: Sys-off mode
477  *	@priority: Handler priority
478  *	@callback: Callback function
479  *	@cb_data: Callback argument
480  *
481  *	Registers resource-managed sys-off handler.
482  *
483  *	Returns zero on success, or error code on failure.
484  */
485 int devm_register_sys_off_handler(struct device *dev,
486 				  enum sys_off_mode mode,
487 				  int priority,
488 				  int (*callback)(struct sys_off_data *data),
489 				  void *cb_data)
490 {
491 	struct sys_off_handler *handler;
492 
493 	handler = register_sys_off_handler(mode, priority, callback, cb_data);
494 	if (IS_ERR(handler))
495 		return PTR_ERR(handler);
496 
497 	return devm_add_action_or_reset(dev, devm_unregister_sys_off_handler,
498 					handler);
499 }
500 EXPORT_SYMBOL_GPL(devm_register_sys_off_handler);
501 
502 /**
503  *	devm_register_power_off_handler - Register power-off handler
504  *	@dev: Device that registers callback
505  *	@callback: Callback function
506  *	@cb_data: Callback's argument
507  *
508  *	Registers resource-managed sys-off handler with a default priority
509  *	and using power-off mode.
510  *
511  *	Returns zero on success, or error code on failure.
512  */
513 int devm_register_power_off_handler(struct device *dev,
514 				    int (*callback)(struct sys_off_data *data),
515 				    void *cb_data)
516 {
517 	return devm_register_sys_off_handler(dev,
518 					     SYS_OFF_MODE_POWER_OFF,
519 					     SYS_OFF_PRIO_DEFAULT,
520 					     callback, cb_data);
521 }
522 EXPORT_SYMBOL_GPL(devm_register_power_off_handler);
523 
524 /**
525  *	devm_register_restart_handler - Register restart handler
526  *	@dev: Device that registers callback
527  *	@callback: Callback function
528  *	@cb_data: Callback's argument
529  *
530  *	Registers resource-managed sys-off handler with a default priority
531  *	and using restart mode.
532  *
533  *	Returns zero on success, or error code on failure.
534  */
535 int devm_register_restart_handler(struct device *dev,
536 				  int (*callback)(struct sys_off_data *data),
537 				  void *cb_data)
538 {
539 	return devm_register_sys_off_handler(dev,
540 					     SYS_OFF_MODE_RESTART,
541 					     SYS_OFF_PRIO_DEFAULT,
542 					     callback, cb_data);
543 }
544 EXPORT_SYMBOL_GPL(devm_register_restart_handler);
545 
546 static struct sys_off_handler *platform_power_off_handler;
547 
548 static int platform_power_off_notify(struct sys_off_data *data)
549 {
550 	void (*platform_power_power_off_cb)(void) = data->cb_data;
551 
552 	platform_power_power_off_cb();
553 
554 	return NOTIFY_DONE;
555 }
556 
557 /**
558  *	register_platform_power_off - Register platform-level power-off callback
559  *	@power_off: Power-off callback
560  *
561  *	Registers power-off callback that will be called as last step
562  *	of the power-off sequence. This callback is expected to be invoked
563  *	for the last resort. Only one platform power-off callback is allowed
564  *	to be registered at a time.
565  *
566  *	Returns zero on success, or error code on failure.
567  */
568 int register_platform_power_off(void (*power_off)(void))
569 {
570 	struct sys_off_handler *handler;
571 
572 	handler = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
573 					   SYS_OFF_PRIO_PLATFORM,
574 					   platform_power_off_notify,
575 					   power_off);
576 	if (IS_ERR(handler))
577 		return PTR_ERR(handler);
578 
579 	platform_power_off_handler = handler;
580 
581 	return 0;
582 }
583 EXPORT_SYMBOL_GPL(register_platform_power_off);
584 
585 /**
586  *	unregister_platform_power_off - Unregister platform-level power-off callback
587  *	@power_off: Power-off callback
588  *
589  *	Unregisters previously registered platform power-off callback.
590  */
591 void unregister_platform_power_off(void (*power_off)(void))
592 {
593 	if (platform_power_off_handler &&
594 	    platform_power_off_handler->cb_data == power_off) {
595 		unregister_sys_off_handler(platform_power_off_handler);
596 		platform_power_off_handler = NULL;
597 	}
598 }
599 EXPORT_SYMBOL_GPL(unregister_platform_power_off);
600 
601 static int legacy_pm_power_off(struct sys_off_data *data)
602 {
603 	if (pm_power_off)
604 		pm_power_off();
605 
606 	return NOTIFY_DONE;
607 }
608 
609 static void do_kernel_power_off_prepare(void)
610 {
611 	blocking_notifier_call_chain(&power_off_prep_handler_list, 0, NULL);
612 }
613 
614 /**
615  *	do_kernel_power_off - Execute kernel power-off handler call chain
616  *
617  *	Expected to be called as last step of the power-off sequence.
618  *
619  *	Powers off the system immediately if a power-off handler function has
620  *	been registered. Otherwise does nothing.
621  */
622 void do_kernel_power_off(void)
623 {
624 	struct sys_off_handler *sys_off = NULL;
625 
626 	/*
627 	 * Register sys-off handlers for legacy PM callback. This allows
628 	 * legacy PM callbacks temporary co-exist with the new sys-off API.
629 	 *
630 	 * TODO: Remove legacy handlers once all legacy PM users will be
631 	 *       switched to the sys-off based APIs.
632 	 */
633 	if (pm_power_off)
634 		sys_off = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
635 						   SYS_OFF_PRIO_DEFAULT,
636 						   legacy_pm_power_off, NULL);
637 
638 	atomic_notifier_call_chain(&power_off_handler_list, 0, NULL);
639 
640 	unregister_sys_off_handler(sys_off);
641 }
642 
643 /**
644  *	kernel_can_power_off - check whether system can be powered off
645  *
646  *	Returns true if power-off handler is registered and system can be
647  *	powered off, false otherwise.
648  */
649 bool kernel_can_power_off(void)
650 {
651 	return !atomic_notifier_call_chain_is_empty(&power_off_handler_list) ||
652 		pm_power_off;
653 }
654 EXPORT_SYMBOL_GPL(kernel_can_power_off);
655 
656 /**
657  *	kernel_power_off - power_off the system
658  *
659  *	Shutdown everything and perform a clean system power_off.
660  */
661 void kernel_power_off(void)
662 {
663 	kernel_shutdown_prepare(SYSTEM_POWER_OFF);
664 	do_kernel_power_off_prepare();
665 	migrate_to_reboot_cpu();
666 	syscore_shutdown();
667 	pr_emerg("Power down\n");
668 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
669 	machine_power_off();
670 }
671 EXPORT_SYMBOL_GPL(kernel_power_off);
672 
673 DEFINE_MUTEX(system_transition_mutex);
674 
675 /*
676  * Reboot system call: for obvious reasons only root may call it,
677  * and even root needs to set up some magic numbers in the registers
678  * so that some mistake won't make this reboot the whole machine.
679  * You can also set the meaning of the ctrl-alt-del-key here.
680  *
681  * reboot doesn't sync: do that yourself before calling this.
682  */
683 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
684 		void __user *, arg)
685 {
686 	struct pid_namespace *pid_ns = task_active_pid_ns(current);
687 	char buffer[256];
688 	int ret = 0;
689 
690 	/* We only trust the superuser with rebooting the system. */
691 	if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
692 		return -EPERM;
693 
694 	/* For safety, we require "magic" arguments. */
695 	if (magic1 != LINUX_REBOOT_MAGIC1 ||
696 			(magic2 != LINUX_REBOOT_MAGIC2 &&
697 			magic2 != LINUX_REBOOT_MAGIC2A &&
698 			magic2 != LINUX_REBOOT_MAGIC2B &&
699 			magic2 != LINUX_REBOOT_MAGIC2C))
700 		return -EINVAL;
701 
702 	/*
703 	 * If pid namespaces are enabled and the current task is in a child
704 	 * pid_namespace, the command is handled by reboot_pid_ns() which will
705 	 * call do_exit().
706 	 */
707 	ret = reboot_pid_ns(pid_ns, cmd);
708 	if (ret)
709 		return ret;
710 
711 	/* Instead of trying to make the power_off code look like
712 	 * halt when pm_power_off is not set do it the easy way.
713 	 */
714 	if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !kernel_can_power_off())
715 		cmd = LINUX_REBOOT_CMD_HALT;
716 
717 	mutex_lock(&system_transition_mutex);
718 	switch (cmd) {
719 	case LINUX_REBOOT_CMD_RESTART:
720 		kernel_restart(NULL);
721 		break;
722 
723 	case LINUX_REBOOT_CMD_CAD_ON:
724 		C_A_D = 1;
725 		break;
726 
727 	case LINUX_REBOOT_CMD_CAD_OFF:
728 		C_A_D = 0;
729 		break;
730 
731 	case LINUX_REBOOT_CMD_HALT:
732 		kernel_halt();
733 		do_exit(0);
734 
735 	case LINUX_REBOOT_CMD_POWER_OFF:
736 		kernel_power_off();
737 		do_exit(0);
738 		break;
739 
740 	case LINUX_REBOOT_CMD_RESTART2:
741 		ret = strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1);
742 		if (ret < 0) {
743 			ret = -EFAULT;
744 			break;
745 		}
746 		buffer[sizeof(buffer) - 1] = '\0';
747 
748 		kernel_restart(buffer);
749 		break;
750 
751 #ifdef CONFIG_KEXEC_CORE
752 	case LINUX_REBOOT_CMD_KEXEC:
753 		ret = kernel_kexec();
754 		break;
755 #endif
756 
757 #ifdef CONFIG_HIBERNATION
758 	case LINUX_REBOOT_CMD_SW_SUSPEND:
759 		ret = hibernate();
760 		break;
761 #endif
762 
763 	default:
764 		ret = -EINVAL;
765 		break;
766 	}
767 	mutex_unlock(&system_transition_mutex);
768 	return ret;
769 }
770 
771 static void deferred_cad(struct work_struct *dummy)
772 {
773 	kernel_restart(NULL);
774 }
775 
776 /*
777  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
778  * As it's called within an interrupt, it may NOT sync: the only choice
779  * is whether to reboot at once, or just ignore the ctrl-alt-del.
780  */
781 void ctrl_alt_del(void)
782 {
783 	static DECLARE_WORK(cad_work, deferred_cad);
784 
785 	if (C_A_D)
786 		schedule_work(&cad_work);
787 	else
788 		kill_cad_pid(SIGINT, 1);
789 }
790 
791 #define POWEROFF_CMD_PATH_LEN  256
792 static char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
793 static const char reboot_cmd[] = "/sbin/reboot";
794 
795 static int run_cmd(const char *cmd)
796 {
797 	char **argv;
798 	static char *envp[] = {
799 		"HOME=/",
800 		"PATH=/sbin:/bin:/usr/sbin:/usr/bin",
801 		NULL
802 	};
803 	int ret;
804 	argv = argv_split(GFP_KERNEL, cmd, NULL);
805 	if (argv) {
806 		ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
807 		argv_free(argv);
808 	} else {
809 		ret = -ENOMEM;
810 	}
811 
812 	return ret;
813 }
814 
815 static int __orderly_reboot(void)
816 {
817 	int ret;
818 
819 	ret = run_cmd(reboot_cmd);
820 
821 	if (ret) {
822 		pr_warn("Failed to start orderly reboot: forcing the issue\n");
823 		emergency_sync();
824 		kernel_restart(NULL);
825 	}
826 
827 	return ret;
828 }
829 
830 static int __orderly_poweroff(bool force)
831 {
832 	int ret;
833 
834 	ret = run_cmd(poweroff_cmd);
835 
836 	if (ret && force) {
837 		pr_warn("Failed to start orderly shutdown: forcing the issue\n");
838 
839 		/*
840 		 * I guess this should try to kick off some daemon to sync and
841 		 * poweroff asap.  Or not even bother syncing if we're doing an
842 		 * emergency shutdown?
843 		 */
844 		emergency_sync();
845 		kernel_power_off();
846 	}
847 
848 	return ret;
849 }
850 
851 static bool poweroff_force;
852 
853 static void poweroff_work_func(struct work_struct *work)
854 {
855 	__orderly_poweroff(poweroff_force);
856 }
857 
858 static DECLARE_WORK(poweroff_work, poweroff_work_func);
859 
860 /**
861  * orderly_poweroff - Trigger an orderly system poweroff
862  * @force: force poweroff if command execution fails
863  *
864  * This may be called from any context to trigger a system shutdown.
865  * If the orderly shutdown fails, it will force an immediate shutdown.
866  */
867 void orderly_poweroff(bool force)
868 {
869 	if (force) /* do not override the pending "true" */
870 		poweroff_force = true;
871 	schedule_work(&poweroff_work);
872 }
873 EXPORT_SYMBOL_GPL(orderly_poweroff);
874 
875 static void reboot_work_func(struct work_struct *work)
876 {
877 	__orderly_reboot();
878 }
879 
880 static DECLARE_WORK(reboot_work, reboot_work_func);
881 
882 /**
883  * orderly_reboot - Trigger an orderly system reboot
884  *
885  * This may be called from any context to trigger a system reboot.
886  * If the orderly reboot fails, it will force an immediate reboot.
887  */
888 void orderly_reboot(void)
889 {
890 	schedule_work(&reboot_work);
891 }
892 EXPORT_SYMBOL_GPL(orderly_reboot);
893 
894 /**
895  * hw_failure_emergency_poweroff_func - emergency poweroff work after a known delay
896  * @work: work_struct associated with the emergency poweroff function
897  *
898  * This function is called in very critical situations to force
899  * a kernel poweroff after a configurable timeout value.
900  */
901 static void hw_failure_emergency_poweroff_func(struct work_struct *work)
902 {
903 	/*
904 	 * We have reached here after the emergency shutdown waiting period has
905 	 * expired. This means orderly_poweroff has not been able to shut off
906 	 * the system for some reason.
907 	 *
908 	 * Try to shut down the system immediately using kernel_power_off
909 	 * if populated
910 	 */
911 	pr_emerg("Hardware protection timed-out. Trying forced poweroff\n");
912 	kernel_power_off();
913 
914 	/*
915 	 * Worst of the worst case trigger emergency restart
916 	 */
917 	pr_emerg("Hardware protection shutdown failed. Trying emergency restart\n");
918 	emergency_restart();
919 }
920 
921 static DECLARE_DELAYED_WORK(hw_failure_emergency_poweroff_work,
922 			    hw_failure_emergency_poweroff_func);
923 
924 /**
925  * hw_failure_emergency_poweroff - Trigger an emergency system poweroff
926  *
927  * This may be called from any critical situation to trigger a system shutdown
928  * after a given period of time. If time is negative this is not scheduled.
929  */
930 static void hw_failure_emergency_poweroff(int poweroff_delay_ms)
931 {
932 	if (poweroff_delay_ms <= 0)
933 		return;
934 	schedule_delayed_work(&hw_failure_emergency_poweroff_work,
935 			      msecs_to_jiffies(poweroff_delay_ms));
936 }
937 
938 /**
939  * hw_protection_shutdown - Trigger an emergency system poweroff
940  *
941  * @reason:		Reason of emergency shutdown to be printed.
942  * @ms_until_forced:	Time to wait for orderly shutdown before tiggering a
943  *			forced shudown. Negative value disables the forced
944  *			shutdown.
945  *
946  * Initiate an emergency system shutdown in order to protect hardware from
947  * further damage. Usage examples include a thermal protection or a voltage or
948  * current regulator failures.
949  * NOTE: The request is ignored if protection shutdown is already pending even
950  * if the previous request has given a large timeout for forced shutdown.
951  * Can be called from any context.
952  */
953 void hw_protection_shutdown(const char *reason, int ms_until_forced)
954 {
955 	static atomic_t allow_proceed = ATOMIC_INIT(1);
956 
957 	pr_emerg("HARDWARE PROTECTION shutdown (%s)\n", reason);
958 
959 	/* Shutdown should be initiated only once. */
960 	if (!atomic_dec_and_test(&allow_proceed))
961 		return;
962 
963 	/*
964 	 * Queue a backup emergency shutdown in the event of
965 	 * orderly_poweroff failure
966 	 */
967 	hw_failure_emergency_poweroff(ms_until_forced);
968 	orderly_poweroff(true);
969 }
970 EXPORT_SYMBOL_GPL(hw_protection_shutdown);
971 
972 static int __init reboot_setup(char *str)
973 {
974 	for (;;) {
975 		enum reboot_mode *mode;
976 
977 		/*
978 		 * Having anything passed on the command line via
979 		 * reboot= will cause us to disable DMI checking
980 		 * below.
981 		 */
982 		reboot_default = 0;
983 
984 		if (!strncmp(str, "panic_", 6)) {
985 			mode = &panic_reboot_mode;
986 			str += 6;
987 		} else {
988 			mode = &reboot_mode;
989 		}
990 
991 		switch (*str) {
992 		case 'w':
993 			*mode = REBOOT_WARM;
994 			break;
995 
996 		case 'c':
997 			*mode = REBOOT_COLD;
998 			break;
999 
1000 		case 'h':
1001 			*mode = REBOOT_HARD;
1002 			break;
1003 
1004 		case 's':
1005 			/*
1006 			 * reboot_cpu is s[mp]#### with #### being the processor
1007 			 * to be used for rebooting. Skip 's' or 'smp' prefix.
1008 			 */
1009 			str += str[1] == 'm' && str[2] == 'p' ? 3 : 1;
1010 
1011 			if (isdigit(str[0])) {
1012 				int cpu = simple_strtoul(str, NULL, 0);
1013 
1014 				if (cpu >= num_possible_cpus()) {
1015 					pr_err("Ignoring the CPU number in reboot= option. "
1016 					"CPU %d exceeds possible cpu number %d\n",
1017 					cpu, num_possible_cpus());
1018 					break;
1019 				}
1020 				reboot_cpu = cpu;
1021 			} else
1022 				*mode = REBOOT_SOFT;
1023 			break;
1024 
1025 		case 'g':
1026 			*mode = REBOOT_GPIO;
1027 			break;
1028 
1029 		case 'b':
1030 		case 'a':
1031 		case 'k':
1032 		case 't':
1033 		case 'e':
1034 		case 'p':
1035 			reboot_type = *str;
1036 			break;
1037 
1038 		case 'f':
1039 			reboot_force = 1;
1040 			break;
1041 		}
1042 
1043 		str = strchr(str, ',');
1044 		if (str)
1045 			str++;
1046 		else
1047 			break;
1048 	}
1049 	return 1;
1050 }
1051 __setup("reboot=", reboot_setup);
1052 
1053 #ifdef CONFIG_SYSFS
1054 
1055 #define REBOOT_COLD_STR		"cold"
1056 #define REBOOT_WARM_STR		"warm"
1057 #define REBOOT_HARD_STR		"hard"
1058 #define REBOOT_SOFT_STR		"soft"
1059 #define REBOOT_GPIO_STR		"gpio"
1060 #define REBOOT_UNDEFINED_STR	"undefined"
1061 
1062 #define BOOT_TRIPLE_STR		"triple"
1063 #define BOOT_KBD_STR		"kbd"
1064 #define BOOT_BIOS_STR		"bios"
1065 #define BOOT_ACPI_STR		"acpi"
1066 #define BOOT_EFI_STR		"efi"
1067 #define BOOT_PCI_STR		"pci"
1068 
1069 static ssize_t mode_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1070 {
1071 	const char *val;
1072 
1073 	switch (reboot_mode) {
1074 	case REBOOT_COLD:
1075 		val = REBOOT_COLD_STR;
1076 		break;
1077 	case REBOOT_WARM:
1078 		val = REBOOT_WARM_STR;
1079 		break;
1080 	case REBOOT_HARD:
1081 		val = REBOOT_HARD_STR;
1082 		break;
1083 	case REBOOT_SOFT:
1084 		val = REBOOT_SOFT_STR;
1085 		break;
1086 	case REBOOT_GPIO:
1087 		val = REBOOT_GPIO_STR;
1088 		break;
1089 	default:
1090 		val = REBOOT_UNDEFINED_STR;
1091 	}
1092 
1093 	return sprintf(buf, "%s\n", val);
1094 }
1095 static ssize_t mode_store(struct kobject *kobj, struct kobj_attribute *attr,
1096 			  const char *buf, size_t count)
1097 {
1098 	if (!capable(CAP_SYS_BOOT))
1099 		return -EPERM;
1100 
1101 	if (!strncmp(buf, REBOOT_COLD_STR, strlen(REBOOT_COLD_STR)))
1102 		reboot_mode = REBOOT_COLD;
1103 	else if (!strncmp(buf, REBOOT_WARM_STR, strlen(REBOOT_WARM_STR)))
1104 		reboot_mode = REBOOT_WARM;
1105 	else if (!strncmp(buf, REBOOT_HARD_STR, strlen(REBOOT_HARD_STR)))
1106 		reboot_mode = REBOOT_HARD;
1107 	else if (!strncmp(buf, REBOOT_SOFT_STR, strlen(REBOOT_SOFT_STR)))
1108 		reboot_mode = REBOOT_SOFT;
1109 	else if (!strncmp(buf, REBOOT_GPIO_STR, strlen(REBOOT_GPIO_STR)))
1110 		reboot_mode = REBOOT_GPIO;
1111 	else
1112 		return -EINVAL;
1113 
1114 	reboot_default = 0;
1115 
1116 	return count;
1117 }
1118 static struct kobj_attribute reboot_mode_attr = __ATTR_RW(mode);
1119 
1120 #ifdef CONFIG_X86
1121 static ssize_t force_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1122 {
1123 	return sprintf(buf, "%d\n", reboot_force);
1124 }
1125 static ssize_t force_store(struct kobject *kobj, struct kobj_attribute *attr,
1126 			  const char *buf, size_t count)
1127 {
1128 	bool res;
1129 
1130 	if (!capable(CAP_SYS_BOOT))
1131 		return -EPERM;
1132 
1133 	if (kstrtobool(buf, &res))
1134 		return -EINVAL;
1135 
1136 	reboot_default = 0;
1137 	reboot_force = res;
1138 
1139 	return count;
1140 }
1141 static struct kobj_attribute reboot_force_attr = __ATTR_RW(force);
1142 
1143 static ssize_t type_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1144 {
1145 	const char *val;
1146 
1147 	switch (reboot_type) {
1148 	case BOOT_TRIPLE:
1149 		val = BOOT_TRIPLE_STR;
1150 		break;
1151 	case BOOT_KBD:
1152 		val = BOOT_KBD_STR;
1153 		break;
1154 	case BOOT_BIOS:
1155 		val = BOOT_BIOS_STR;
1156 		break;
1157 	case BOOT_ACPI:
1158 		val = BOOT_ACPI_STR;
1159 		break;
1160 	case BOOT_EFI:
1161 		val = BOOT_EFI_STR;
1162 		break;
1163 	case BOOT_CF9_FORCE:
1164 		val = BOOT_PCI_STR;
1165 		break;
1166 	default:
1167 		val = REBOOT_UNDEFINED_STR;
1168 	}
1169 
1170 	return sprintf(buf, "%s\n", val);
1171 }
1172 static ssize_t type_store(struct kobject *kobj, struct kobj_attribute *attr,
1173 			  const char *buf, size_t count)
1174 {
1175 	if (!capable(CAP_SYS_BOOT))
1176 		return -EPERM;
1177 
1178 	if (!strncmp(buf, BOOT_TRIPLE_STR, strlen(BOOT_TRIPLE_STR)))
1179 		reboot_type = BOOT_TRIPLE;
1180 	else if (!strncmp(buf, BOOT_KBD_STR, strlen(BOOT_KBD_STR)))
1181 		reboot_type = BOOT_KBD;
1182 	else if (!strncmp(buf, BOOT_BIOS_STR, strlen(BOOT_BIOS_STR)))
1183 		reboot_type = BOOT_BIOS;
1184 	else if (!strncmp(buf, BOOT_ACPI_STR, strlen(BOOT_ACPI_STR)))
1185 		reboot_type = BOOT_ACPI;
1186 	else if (!strncmp(buf, BOOT_EFI_STR, strlen(BOOT_EFI_STR)))
1187 		reboot_type = BOOT_EFI;
1188 	else if (!strncmp(buf, BOOT_PCI_STR, strlen(BOOT_PCI_STR)))
1189 		reboot_type = BOOT_CF9_FORCE;
1190 	else
1191 		return -EINVAL;
1192 
1193 	reboot_default = 0;
1194 
1195 	return count;
1196 }
1197 static struct kobj_attribute reboot_type_attr = __ATTR_RW(type);
1198 #endif
1199 
1200 #ifdef CONFIG_SMP
1201 static ssize_t cpu_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1202 {
1203 	return sprintf(buf, "%d\n", reboot_cpu);
1204 }
1205 static ssize_t cpu_store(struct kobject *kobj, struct kobj_attribute *attr,
1206 			  const char *buf, size_t count)
1207 {
1208 	unsigned int cpunum;
1209 	int rc;
1210 
1211 	if (!capable(CAP_SYS_BOOT))
1212 		return -EPERM;
1213 
1214 	rc = kstrtouint(buf, 0, &cpunum);
1215 
1216 	if (rc)
1217 		return rc;
1218 
1219 	if (cpunum >= num_possible_cpus())
1220 		return -ERANGE;
1221 
1222 	reboot_default = 0;
1223 	reboot_cpu = cpunum;
1224 
1225 	return count;
1226 }
1227 static struct kobj_attribute reboot_cpu_attr = __ATTR_RW(cpu);
1228 #endif
1229 
1230 static struct attribute *reboot_attrs[] = {
1231 	&reboot_mode_attr.attr,
1232 #ifdef CONFIG_X86
1233 	&reboot_force_attr.attr,
1234 	&reboot_type_attr.attr,
1235 #endif
1236 #ifdef CONFIG_SMP
1237 	&reboot_cpu_attr.attr,
1238 #endif
1239 	NULL,
1240 };
1241 
1242 #ifdef CONFIG_SYSCTL
1243 static struct ctl_table kern_reboot_table[] = {
1244 	{
1245 		.procname       = "poweroff_cmd",
1246 		.data           = &poweroff_cmd,
1247 		.maxlen         = POWEROFF_CMD_PATH_LEN,
1248 		.mode           = 0644,
1249 		.proc_handler   = proc_dostring,
1250 	},
1251 	{
1252 		.procname       = "ctrl-alt-del",
1253 		.data           = &C_A_D,
1254 		.maxlen         = sizeof(int),
1255 		.mode           = 0644,
1256 		.proc_handler   = proc_dointvec,
1257 	},
1258 	{ }
1259 };
1260 
1261 static void __init kernel_reboot_sysctls_init(void)
1262 {
1263 	register_sysctl_init("kernel", kern_reboot_table);
1264 }
1265 #else
1266 #define kernel_reboot_sysctls_init() do { } while (0)
1267 #endif /* CONFIG_SYSCTL */
1268 
1269 static const struct attribute_group reboot_attr_group = {
1270 	.attrs = reboot_attrs,
1271 };
1272 
1273 static int __init reboot_ksysfs_init(void)
1274 {
1275 	struct kobject *reboot_kobj;
1276 	int ret;
1277 
1278 	reboot_kobj = kobject_create_and_add("reboot", kernel_kobj);
1279 	if (!reboot_kobj)
1280 		return -ENOMEM;
1281 
1282 	ret = sysfs_create_group(reboot_kobj, &reboot_attr_group);
1283 	if (ret) {
1284 		kobject_put(reboot_kobj);
1285 		return ret;
1286 	}
1287 
1288 	kernel_reboot_sysctls_init();
1289 
1290 	return 0;
1291 }
1292 late_initcall(reboot_ksysfs_init);
1293 
1294 #endif
1295